Video delivery is typically performed by a set of servers on which video in the form of chunks is stored. When a client requests video, the client may connect to the servers and may obtain the chunks from the servers until the video is completed. The client may open multiple connections to substantially all servers that contain different chunks of a video. Obtaining video through multiple connections is not efficient, and generally utilizes significant network resources.